The promotion of biofuels as energy for transportation in the industrialized countries is mainly driven by the perspective of oil depletion, the concerns about energy security and global warming. However due to sustainability constraints, biofuels will replace only 10 to 15% of fossil liquid fuels in the transport sector. Several governments have defined a minimum target of GHG emissions reduction for those biofuels that will be eligible to public incentives, for example a 35% emissions reduction in case of biofuels in Members States of the European Union. This article points out the significant biases in estimating GHG balances of biofuels stemming from modelling choices about system definition and boundaries, functional unit, reference systems and allocation methods. The extent to which these choices influence the results is investigated. After performing a comparison and constructive criticism of various modelling choices, the LCA of wheat-to-bioethanol is used as an illustrative case where bioethanol is blended with gasoline at various percentages (E5, E10 and E85). The performance of these substitution options is evaluated as well. The results show a large difference in the reduction of the GHG emissions with a high sensitivity to the following factors: the method used to allocate the impacts between the co-products, the type of reference systems, the choice of the functional unit and the type of blend. The authors come out with some recommendations for basing the estimation of energy and GHG balances of biofuels on principles such as transparency, consistency and accuracy.